Econometrics Research Papers

GAZİ ÜNİVERSİTESİ, EĞİTİM FAKÜLTESİ ÖĞRENCİLERİNİN ÜNİVERSİTE YAŞAMINA UYUM DURUMLARI

by Hatice Bekir

2005, Gazi Üniversitesi Kastamonu …

descriptionView Paper arrow_downwardDownload

MODUL PRAKTIKUM EKONOMETRIKA

by Rezzy Eko Caraka

descriptionView Paper arrow_downwardDownload

Pengantar Ekonometrika (Teori dan Praktek)

by Arif Rahman Hakim

2014

Ekonometrika dapat diartikan secara harfiah sebagai sebuah ukuran ekonomi. Secara definitif, ekonometrika, diartikan sebagai suatu ilmu yang mempelajari analisis kuantitatif dari fenomena ekonomi. Alternatif definisi lainnya merupakan... more

descriptionView Paper arrow_downwardDownload

Basic econometrics using STATA

by MUHAMMAD ZUBAIR CHISHTI

descriptionView Paper arrow_downwardDownload

The real effects of monetary policy in China: An empirical analysis

by Quỳnh Như Ôn

This paper examines whether institutional changes have affected the interaction between the real economy and monetary policy in China. We find evidence that structural changes in the financial and real sectors over the period of our study... more

In this case we observe a significant positive response to output in the latter period compared to the insignificant effects in the first period. The interpretation is the same as we observed with system 3. We now turn to consider the systems with total bank loans as a policy variable. Note that we find evidence of instability 1984.10 in our estimated relationships and hence start our period from 1985.1.

The clear effect is that total loan shocks have smaller and less persistent effects on output in the second sub-period. This would indicate a change in the way banks are allocating credit with possibly more emphasis on lending to SOEs which are in financial distress and hence a much smaller impact of a shock on output.

First sub-period Hi fo summarise the results of system | to 6, all the models have confirmed that the central bank lending rate was not a significant policy instrument in the pre-1990 period. In contrast bank lending has been seen to have a significant effect on SOE output. There is evidence that the effect holds for collective and private enterprises in the 1984 to noting that fixed asset investment was a key obj 990 period and we explain this by ective of bank lending in the earlier sub-period, but that increasingly it was used to maintain loss-making SOEs in the later period.*° The CBLR has been seen to have a significant effect on co lective and private output but with the opposite effect to that conventionally expected. The reason we have given for this is that increases in the CBLR have the effect of giving more access to that can then raise output.*' We do not find SO and hence total output rises as a consequence. bank funds by collective and private enterprises E activity significantly affected by interest rates

Response to Generalized One S.D. Innovations + 2 S.E. First sub-period

Response to Generalized One S.D. Innovations + 2 S.E.

Examining the results above we find evidence consistent with our previous analysis. For the system with credit we find that there is a break very early in the time period. Therefore we run our estimation period from 1985.01 and only report for one break which is the optimum.

Unit root tests: KPSS, Augmented Dickey—Fuller (ADF) test and Phillips—Perron (PP) test Unit root tests include constant; for ADF the optimal lag length is chosen by AIC criterion; for KPSS and PP the bandwidtl is chosen using Newey—West method and spectral estimation uses Bartlett kernel. *Signifies statistically significant at 5%, **Significant at 1% level. * The null hypothesis for KPSS is that the variable is stationary, for ADF and PP that the variable has a unit root. Table 1

Appendix B. Systems with total and sectoral output, price, policy variable, money In this appendix we report the result from models that include both total output and sectoral output. This allows us to pick up the separate effects of output and sectoral output on policy.

descriptionView Paper arrow_downwardDownload

Identification of Causal Effects Using Instrumental Variables: Rejoinder

by Guido Imbens

1996, Journal of the American Statistical Association

Table 1. Causal Effect of Z on Y, Y;(1, Dj(1)) — Yj(0, D;(0)), for the Population of Units Classified by D;(0) and D;(1) the exclusion restriction and the monotonicity assumption, it is randomly (or ignorably) assigned, and SUTVA holds (i.e., if Assumptions 1-5 hold). SUTVA and the exclusion restriction are sufficient to es- tablish a fundamental relationship between the intention-to- treat effects of Z on Y and D and the causal effect of D on Y at the unit level:

2 Determined by lottery number cutoff: RSN 195 for men born in 1950, and RSN 125 for men born in 1951. © From California and Pennsylvania administrative records, all deaths 1974-1983. Data sources and methods documented by Hearst et al. (1986). Note: Sample sizes differ from Hearst et al., because non-U.S.-born are included to match SIPP data in the last column. © The mortality figures are tabulated from the data set analyzed by Hearst et al. (1986). ¢ The estimated population at risk is from the author's tabulation of 1970 census data. Estimates by draft-eligibility status are computed assuming a uniform distribution of lottery numbers. Standard errors are given in parentheses. ® These figures are taken from Angrist (1990), table 2, and were tabulated using a special version of the SIPP that has been matched to indicators of draft eligibility. Note that probabilities estimated using the SIPP are for the entire country and do not take account of morality. The impact of mortality on differences in the probability of being a veteran by eligibility status is small enough to have only trivial consequences for the estimation. f The standard errors, following econometric practice (e.g. Imbens and Angrist 1994), were calculated based on a normal approximation to the sampling distribution of the ratio of the difference in estimated probability of death/suicide and the difference in estimated probability of serving. We assume independence of numerator and denominator because they were calculated from different data sets. Pooled estimates show a statistically significant increase in risk at conventional significance levels (e.g., Hearst, Newman, Hulley 1986). Table 2. Data on Civilian Mortality for White Men Born in 1950 and 1951

descriptionView Paper arrow_downwardDownload

Journal of Emerging Issues in Economics, Finance and Banking (JEIEFB) Factors Affecting the Price of Housing in Malaysia

by Shasha Saffari

The main purpose of this study is to measure the relationship between macroeconomic variables and the housing price. This paper examines empirically whether the increasing trend in the Malaysian housing price is related to changes in the... more

housing price to the length of time in which the house was built with the fluctuation; however, this

descriptionView Paper arrow_downwardDownload

Statistical Foundations for Econometric Techniques

by Asad Zaman

1996

1.2. Projection: the simplest case A series of observations on a variable can be regarded as a vector. Say we observe y: for t = 1,2,...,T. Then we can take y = [y1, y2,...,yr]’ asa T X 1 vector in R?. Let x € R" be some other vector and a € R a scalar. The set of vectors ax is the one- dimensional vector space generated by x. This vector space contains all vectors which can be “explained” by x. The projection of y on x is a vector of the type ax which comes as close as possible to y. In other words, it is the best explanation of y which is possible using x. It is easy to calculate the projection of y on x using calculus. We want to find a which minimizes the distance between ax and y. This distance is d(a) = ||y — ax||? = Il yl|? — 2ay’x + a*||x||?. Minimizing this quadratic with respect to « yields

The top K x K block of the information matrix is X’X/o7. Since the off-diagonals are zero because u(y — XB) = t€ = 0, the top K x K block of the inverse of the information matrix will be 07(X’X)~!, which is precisely the covariance of the ML estimator. It follows that B is MVU. As for the bottom (K +1), (K+1) entry, since E]| y—XB||? = it follows that Leiixi1 = —T/(20*) + T/o* = T/(20*). Since the information matrix is block diagonal, the CRLB for variance of unbiased estimators of 07 is 204/T. Now SSR = lly - Xl? is distributed as 0? y7_x so that Var(SSR) = 204(T — K). It follows that le ||? = To?. ge ee e ae a at aie. A

Figure 4.1. Coordinate risks of James-Stein estimate.

Ficure 5.1. Breakdown of LAD with X-outlier.

Figure 5.2. LMS and OLS fit to brains data: adapted from Rousseeuw and Leroy [167]. Robust regression and outlier detection. Reprinted by permission of John Wiley & Sons, Inc. Because the LMS finds a good fit for a majority of the data, rather than finding a mediocre fit for all of the data, the contrasts between good and bad fitting points are brought much more sharply into focus. This idea can be clarified by considering the “Exact Fit Property” of LMS.

Ficure 5.3. The leverage of giant stars: adapted from Rousseeuw and Yohai [170] with permission of Springer-Verlag.

a neighborhood of the null (where it is more or less impossible to achieve) the LMP test sacrifices good performance at distant points. However, for practical purposes it is more important to detect a violation of the null hypothesis when the violation is large (i.e., the alternative is distant) than when the violation is small. Tests which maximize power over suitably distant alternatives are also preferable for this reason.

Ficure 6.2. Berenblut-Webb test shortcomings.

Figure 6.3. Shortcomings versus maximum power.

Ficure 6.4. LM, Wald, and LR shortcomings. Algorithms

If we assume EV so that the variances are equal across the regressions, an optimal test for the hypothesis EM is the usual F test. We will refer to each of the separate regression equations as a regime in the sequel. Below we present the test in a form which allows for the possibility that some of the regimes may have insufficient observations to allow an estimate B ; of B;. For this case, let y; be the projection of y; onto the column space of the matrix X;. This is always well defined and unique, even though the estimates B j= (X;X;)"'X;y; are not well defined in the case that T; < K. Let Be be the estimate of 6 under the constraint EM that all regimes have identical coefficients. When both EM and EV hold, all of the data can be aggregated into one regression as follows. Define

Every kth-order autoregressive dynamic model such as (11.20) can be represented as a first-order vector autoregressive model. This representation permits convenient deriva- tion of the conditions for stationarity, and has other uses as well. Define the following vectors and matrices a first-order vector autoregressive model. This representation permits convenient deriva- For a proof, see Lai and Wei [118]. This condition is also close to being necessary. La and Wei produce an example with stochastic regressors where this condition is slightly violated and the strong consistency fails. Use of this result permits proof of consistency of OLS in dynamic models under the most general conditions currently available. The details of the proof are somewhat involved and will not be presented here. To apply thi: result to particular dynamic models, we need to be able to estimate Amax and Amin for the matrix X’X arising in dynamic models. Various theorems permitting one to estimate these eigenvalues and apply these results are developed by Lai and Wei [119]. We will now present some of these results. Many physical and economic systems satisfy the condition that the effects of the distant past should be negligible on the present. On the present equation, this translates to the requirement that the term A¥*! Y;_(x41) should be small for large K. This is equivalent

Ficure 12.1. Convergent functions with divergent maxima. A sequence f,(@) is said to converge pointwise to a limit function f if for each Olimy—co fr(O) = f (0). First we note that pointwise convergence is not enough for the kind of result that we need. This can be seen in Figure 12.1(a), which plots the functions fr(x) = e-*+(x—n), where ¢ is the normal density. For fixed x as n approaches infinity $(x — n) goes to zero, so that f,(x) converges to fi.(x) = e-*. However, it is easily seen that f,(x) has a maximum at x = n, while f..(x) has a maximum at x = 0 for x = 0. The difficulty with pointwise convergence, as can be seen from Figure 12.1(a), is that the maximum gap between the functions f,, and the limit function f need not decrease with large n. This means that the shape of f,, for large n does not necessarily conform to the shape of the limiting function. We can strengthen the notion of convergence to get the desired result as follows.

Ficure 12.2. Empirical and actual normal CDE.

14.5.2. Use of theory to guide the bootstrap

also 6? is also close to the true value. Thus the bootstrap will work well for some types of samples and poorly for others. In studying the behavior of the bootstrap, it is essential to do repeated experiments to determine how reliable the bootstrap is in a variety of samples. With this basic principle understood, we go on to discuss the many lessons we can learn from the figures.

Ficure 16.1. Batting average forecasts by ML, EB, and mean ML.

Proof. The Laplace transform g = £, is calculated as follows: This property is equivalent to uniqueness. For a, b € R", let us write a < b to indicate that the inequality holds in each coordinate. Suppose a < b and suppose that the Laplace aah of two functions f(x) and g(x) is the same for all 6 such that a < @ < b. Define = f(x) — g(x) and note that £,(0) = 0 for all 6 such that a < @ < b. Now the orperts above implies that z(x) = 0 for all x, and hence f is the same function as g.

TaBLe 5.1. Brains data and standardized residuals. Note: Adapted from Rousseeuw and Leroy [167]. Robust Regression and Outlier Detection. Reprinted by permission of John Wiley & Sons, Inc.

TABLE 9.1. Winning speed at Indy 500. the mean, and T>, T3, Ty so that each T; is orthogonal to the previous regressors (see Exercise 9.10(1)). A regression of W; on the vector of ones, and these four regressors yields the following result:

TABLE 9.2. Forecast performance of consumption models. Applications: F and related tests

TABLE 16.1. 1970 major league baseball batting averages.

TABLE 16.2. Recessions and expansions 1854 to 1990.

TABLE 16.3. Batting average data for Thurman Munson. a aaa nae” | Aka cai The MOM is not necessarily an efficient estimator, since the prior variance is esti- mated as a simple average of the estimates of variance from each OMe, It would be more efficient to use a weighted average, with weights being the precisions. To get these preci- sions, we will need the fourth moments of the marginal density of D;. These are in fact derived and used by Hoadley [96] in developing the formulae for the QMP. An alternative, perhaps simpler, route to getting efficient estimates is to use ML. For example, we could solve (16.35) for fio and numerically maximize the marginal density m(D,) derived above (which is in fact a negative Binomial) with respect to the parameter e, after substituting d = fice. By this means we get a one parameter maximization problem which is typically substantially easier to solve than a two-dimensional one.

TABLE 16.5. Defects in keys of telephone sets-Shreveport.

descriptionView Paper arrow_downwardDownload

THE IMPACT OF MACROECONOMIC VARIABLES ON THE PROFITABILITY OF LISTED COMMERCIAL BANKS IN PAKISTAN

by Chief Editor

ased on vital contribution of the commercial banks to economic progress of Pakistan, this study investigates the impact of macroeconomic variables on profitability of public limited commercial banks in Pakistan for years method is used to... more

Table 1. Dependent Variables and Their Assessment Present study employs technique of Descriptive statistics, Pearson Correlation and Pooled Ordinary Least Squares (POLS) Regression. POLS method is used because literature suggests that it is a valid method where variables show stable relationship across the banks (Gul, Irshad, & Zaman, 2011). Table 1 gives the details of dependent variables and Table 2 shows the details of independent variables respectively.

Table 2. Independent Variables and Their A ssessment ‘Real interest rate for years 2001, 2002 and 2003 were calculated using average lending rate for a year adjusted with inflation

Table 3. Descriptive Statistics for Variables Included in Study According to table 3, on average the total return on assets of listed banks for past decade was 0.7% where as the ROE was 22.6%. ROE between 15% -20% is considered favorable (Fraker, 2006). Also the banks selected in this study vary due to their size and level of establishment. The average EM ratio of banks was about 24% for years under study. Table also depicts that Pakistan’s average consumer price inflation rate has been 9% during this time. As far as real interest rate is concerned, in 10 years, banks have been charging around 10% rate of real interest. And on average the growth rate of Pakistan’s economy is 4.6% during 2002-2011.

* Significant at 1% level, ** Significant at 5% level and (NS) = Not significant Below is the regression equation for first model, which shows the three external factors influencing ROA. Table 5 gives the details of empirical results of model 1. Results of Model 1

Table 6. Empirical Findings Where Dependent Variable: ROE According to table 6 the value of R? in both steps is around 3% showing very less contribution of selected macroeconomic variables towards ROE. Inflation rate exerts a negative impact on ROE ratio with p- value<0.05 indicating its significance at 5% level. GDP in both cases remains an insignificant factor affecting ROE of banks but it has affected ROE negatively, whereas interest rate is observed to have a strong positive relationship with ROE at 1% level of significance.

* Significant at 1% level, ** Significant at 5% level and (NS) = Not significant Table 7. Empirical Findings Where Dependent Variable: EM

List of Commercial Banks Included in Study

descriptionView Paper arrow_downwardDownload

Introduction to Bayesian Econometrics Introduction to Bayesian Econometrics

by Jose Perea Arango

This concise textbook is an introduction to econometrics from the Bayesian viewpoint. It begins with an explanation of the basic ideas of subjective probability and shows how subjective probabilities must obey the usual rules of... more

Figure 2.1. Beta distributions for various values of a and f.

Although the preceding discussion shows that the beta prior is a “natural” prior for Bernoulli data and that the choice of the two parameters in the beta prior can capture a wide variety of prior beliefs, it is important to note that it is not necessary to adopt a beta prior if no combination of parameters can approximate the prior you wish to specify. Beta priors, for example, do not easily accommodate bimodal distributions. We describe methods later in the book that can approximate the posterior distribution for any specified prior, even if the prior information does not lead to a posterior distribution of a standard form.

If the prior distribution of 6) is independent of 6,, that is, 77(62|0;) = 7 (62), then (@2|y) = (62), implying that knowledge of y does not modify beliefs about 6. The discussion of identification to this point has been based on the specificatior of the likelihood function, what we might call “identification through the data, but the Bayesian approach also utilizes a prior distribution. Consider the likelihooc function f(y|6,, 62) = f(y). It is clear that the data have no information abou 7, when 6, is given, but what can be said about the posterior distribution z (62|y)’ Although we might expect that it is equal to 2(@2) since the data contain n information about 62, consider the following calculation:

Since the second square-bracketed term does not depend on n, its importance may be neglected for large n. The first term in the first square bracket is the logarithm of the likelihood ratio. It will tend to become large if M, is the true model and small if Mp) is true. The second term shows that the log Bayes factor penalizes models with larger numbers of parameters, where the penalty is log(7) times the difference in the number of parameters divided by two. where d; is the dimension of 6;; that is, the number of parameters in M;. We can now approximate the logarithm of the Bayes factor for comparing models 1 and 2:

Given a sample of size n and assuming that the u; are independent, we can write the probability density for the observed sample, which is also the likelihood function for the unknown parameters, as To derive this expression, we have used the normality of u; and the transformation of random variables from u; to y; based on y; — x/6 = u;, which has a Jacobian of one. In the last line, we absorbed the (277)~"”? into a proportionality constant. This is done to focus attention on the important terms in the expression and is possible because it is a multiplicative term that does not contain the unknown parameters 6 and o?. Here and in the following, we follow the convention of usually not explicitly including the covariates X in the conditioning set of the posterior distribution.

Figure 4.1. Prior and posterior distributions for coefficient of union membership.

Figure 5.1. Target and proposal density to sample from Beta(3, 3).

Figure 5.2. Target and proposal density to sample from N(0, 1).

Figure 5.3. Simulation results for Z = XY, where X ~ Beta(3, 3) and Y ~ Beta(5, 3).

be introduced. To see the problem, consider the Markov chain defined by thi following transition probabilities:

To salvage a counterpart of Theorems 6.1 and 6.2, the concept of recurrence is needed.

We now verify that the invariant distribution is Gaussian with those parameters: the last expression is that of a N(0, o~/(1 — 6°)) distribution as was to be shown.

Figure 7.1. Simulation results for MH sampling of Beta(3, 4) with U(O, 1) proposal.

Figure 8.1. Posterior distributions of By and 0”, Gaussian errors.

Figure 8.2. Autocorrelations of By and o”, Gaussian errors.

Figure 8.3. Posterior distributions of By and o”, Student-t errors. The Chib method for computing the marginal likelihood for this problem is ased on the identity

Figure 8.4. Autocorrelations of By and 7, Student-t errors. 8.2 Limited Dependent Variables This section takes up several models that have the usual normal linear regression form,

Figure 8.5. Posterior distributions of 8: Tobit model, Mroz data.

Figure 8.6. Predictive distributions of hours worked: Tobit model, Mroz data. Left: women with young children; right: women with older children.

We generated 10,000 draws from the joint posterior distribution and analyzed the last 9,000. The summary statistics in Table 8.2 and Figure 8.7 point to positive effects for both covariates. The coefficient of weekly wage is determined rather precisely. For B92, consider a household that earns $2,000 weekly, in which the head is not a college graduate. We assume a probability of 0.8 that such a household has a com- puter. Setting 8; = Bo1, we solve fe 2816+2B 00 o(t) dt = 0.8 and find Bo2 = 1.0616. We further assume it very likely that the probability is at least 0.6 that such a household owns a computer. Accordingly, we solve Jon LARIG+2P p(t) dt = 0.6 to find 62 = .7674, and then find Bo.22 = 0.0225. To specify a prior for 63, we assume the probability that a household with zero weekly wage is 0.4 for a household with a college degree and have previously assumed the probability is 0.1 for a household without a college degree. Computations similar to those mentioned before imply Bo3 = 1.0283 and Bop 33 = 0.2752. To summarize, effects for both covariates. The coefficient of weekly wage is determined rather

Figure 8.7. Posterior distributions of 8: computer ownership example, probit model

Figure 8.8. Posterior distributions of 6: computer ownership example, logit model.

It is then straightforward to determine the conditional distribution, With these assumptions, the posterior distribution is We next specify a conditionally conjugate prior for this model. The regression coefficients are assumed to have a Gaussian prior, 8B ~ Nx(Bo, Bo). For the pre- cision matrix ©~', we assume a generalization of the gamma distribution, the Wishart distribution described in Section A.1.16: 2~! ~ Ws(v9, Ro). (We could have equivalently assumed an inverted Wishart distribution for the covariance matrix ».)

ranging in absolute value from 0.029 to 0.599. The 6 values for Mz show compa- rable variation. The marginal likelihood values are —708.60 for M,, —593.01 for M2, and —651.75 for M3. By this criterion, the data strongly favor Mz, a model that allows each firm to have a different set of regression coefficients, but imposes zero contemporaneous correlations. Another model that might be examined is one in that permits different regression coefficients and different variances for each firm. This possibility is included as an exercise.

whereas M> includes only one variance. The latter model therefore entails 14 fewer parameters than does the former. Although we might expect the likelihood ratio to favor the model with more parameters, M , the effect from this source was apparently not enough to offset the larger number of parameters. In contrast, M3, which contains only four parameters, is decisively beaten by M2 with its 16 parameters.

The regression parameter 6 can therefore be simulated from its conditional distri- bution in a Gibbs step. With ones on the main diagonal, & is a correlation matrix, and, because it is symmetric, there are J(J — 1)/2 unique unknown o;;. We denote these by Oo = (012, 013,---,07-1,7). In addition, the positive definiteness of © imposes restrictions ono.

Figure 9.3. Posterior distributions of By and mean(b2).

The derivation of z(o7?|y, 6, ¢) is also straightforward on the assumption that the prior distribution for h, = 1/0? is G(ao/2, 59/2); then h, | y, B, 6 ~ G(a,/2, 6, /2), where The conditional posterior distributions of 6 and o? can therefore be simulated by Gibbs steps, but that of @ requires an MH step. The joint posterior distribution of the parameters in a form that is convenient for imulating 6 and o? is

Draws of ¢ from that distribution are made until one is found that is in the stationat region; it is then subjected to the usual MH acceptance criterion. where S¢ is the region in which the process is stationary. This distribution is clearly nonstandard, but it can be sampled with an MH algorithm after specifying the prior distribution of ¢; for example, 6 ~ N,(¢o, ®o). A possible proposal generating density is the distribution obtained by multiplying 2 (@) by the terms involving y;, t > p, in (10.4); that distribution is N(¢, &), where

where T; = >, s; and 7) = T — T;. By assuming the usual conditional conjugate distributions for jzo and o? and a truncated normal distribution for /11, it is straight- forward to devise a sampler for jz and o7. Moreover, if we are given S and beta prior distributions for a and b

Figure 10.1. Probability of recession. The model discussed in this section assumes that regression coefficients evolve ran- domly through time. It is written in state-space form and consists of the observation

This algorithm must be modified for large KT because By is the inverse of ¢ KT x KT matrix that may be too large to invert rapidly and accurately. The matrix inversion can be avoided by sampling £, with the same type of recursive scheme utilized in sampling for s, in Section 10.2. KT x KT matrix that may be too large to invert rapidly and accurately. The matrix

It is now straightforward to derive the conditional posterior distributions for specifying a Gibbs algorithm: Incidental truncation arises when the response variable y; is not observed for all units, and whether it is observed depends on the value of a “selection” variable s;. Part or all of the covariate data x; are observed for all sample units. A well-known example of this situation is a model designed to explain the wage rate of married women on the basis of demographic and economic variables, where no wage rate is observed for women who do not work in the time period of observation, but part or all of the demographic and economic variables may be observed. The factors determining the decision to work, which is the binary variable s; = 1 if individual i works and 0 otherwise, may not be independent of the wage rate, so that including in the sample only those women who work is likely to result in a poor estimate of the effects of the covariates of interest. As another example, s; may be a Tobit-type variable rather than a binary variable, where y; is observed when s; > 0. In the

Figure 11.1. Selected coefficients: incidental truncation model, Mroz data. convenient reference is Heckman and Vytlacil (2005), which includes many ref- erences to the literature and definitions of other treatment effects. The Bayesian specification in this section is based on Chib (in press-a), which also treats more general cases. One feature of Chib’s approach is that it does not model the corre- lation between u; and v;. This is desirable because yj and y;,; cannot be observed for the same individual, which implies that there is no sample information that can identify those correlations. Poirier and Tobias (2003) argue that the positive def- initeness constraint on the joint distribution of (uo;, 41;, v;) provides information about the unidentified covariance between uo; and u,; and present an algorithm to sample the regression coefficients and the 3 x 3 covariance matrix.

We can now show that the principle of coherency restricts the value of p you set. If p < 0, your opponent, by choosing S < 0, will inflict a loss (a negative gain) on you whether or not A occurs. By coherency, therefore, p > 0. Similarly, if you set p > 1, your opponent can set S > 0, and you are again sure to lose. Axiom | is therefore implied by the principle of coherency. Consider a simple bet on or against A, where you have set the price of a ticket at p and you are holding S tickets for which you have paid pS; your opponent has chosen S. (Remember that S > 0 means that you are betting on A and S$ <0 means you are betting against A.) If A occurs, you pay pS and collect S. If A does not occur, you collect pS. Verify that these results are valid for both positive and negative values of S. We summarize your gains in the following table, where the rows denote disjoint events and cover all possible outcomes:

To examine the subjective assignment of P(A; U Az) if A; and Az are disjoint, consider the table of gains when you set prices pj, p2, p3, and your opponent chooses S$}, Sz, and $3, respectively, on the events A,, Az, and (A; UA2). A positive (negative) S; again means that you are betting on (against) A,, and the same for S> and $3. As before, the events listed in the first column are a disjoint set that cover all possible outcomes. Verify that the three rows in the second column of the table express your winnings (or your losses if the values are negative) as a function of the values of p;, p2, and p3 that you choose and the values of S,, S2, and $3 that your opponent chooses:

Since the third and fourth payoffs are identical, we can consider only the three distinct equations. As above, to prevent your opponent from being able to win regardless of the outcome you will need to set Since the third and fourth payoffs are identical, we can consider only the three following payoff table in which the rows cover all possible outcomes and are disjoint:

Table 4.1. By as a Function of Hyperparameters Born and Resa 4.8 Sensitivity and Robustness

Table 8.1. Summary of Posterior Distribution: Tobit Model, Mroz Data. set a large variance of 2,000. Finally, consider a husband’s wage increasing from $5 to $10 per hour. We assume this change reduces labor input by 500 hours, so that Bos = —5O. The variance is again set at 2,000 to reflect uncertainty. For o? we assume a standard deviation of 1,200 yielding E(a*) = 1,400,000, and Var(o”) = 100,000. Since the parameters are a/2 and 6/2, the result in (A.2) is modified to a = 2[E(o*)}° / Var(o”) and 6 = 2E (o7)/ Var(o”). Given the aforementioned numbers, we have a = 39,200,000 and 6 = 28.

Table 8.2. Summary of Posterior Distribution: Probit Model, Computer Example.

Table 8.3. Summary of Posterior Distribution: Logit Model, Computer Example.

Table 9.1. Summary of Posterior Distribution of Br: Grunfeld Data, SUR Model.

Table 9.2. Summary of Posterior Distribution of Bc: Grunfeld Data, SUR Model.

where x;; and 6; are K; x 1, and let K = }° K;. The observed choices are the yj; which are related to the latent data through

Table 9.4. Summary of Prior and Posterior Distributions of B and 02: Rubinfeld Data.

Table 9.5. Summary of Posterior Distribution: Panel Data Model, Vella—Verbeek Data. We conclude with an example based on the Vella—Verbeek data discussed in Sections 4.4 and 8.1.1. The intercept and the experience variable are assigned to W; (random effects), and the remaining covariates, including the union membership dummy variable, are placed in X; (fixed effects). The results are summarized in Table 9.5 and Figure 9.3. The coefficient on the union dummy variable based on all 8 years of data is somewhat smaller than that based on only the 1987 data. Mean(b>) indicates the mean of the n values of b;2, the random effect of experience. On average, a year of experience adds about 6% to wage, but the likely effect varies between 3.85 and 7.82%.

Table 10.1. Summary of Posterior Distribution: AR(1) Errors. suggests a moderate amount of autocorrelation. To check this, we ran the same specification without an autocorrelation parameter, where the priors are taken from a regression on the training-sample data, assuming independent errors. The log, of the marginal likelihood for the AR(1) model is —89.028 compared to —94.543 for the model that assumes independent errors, which is a strong evidence in favor of the autocorrelated model.

Table 10.2. Parameter Estimates for GDP Data. not going into recession. The figure suggests a fairly clear-cut distinction between recessionary and nonrecessionary periods since only a few of the 175 observations are around 0.5.

Table 10.3. Summary of Posterior Distribution: Time Varyin; Parameter Model.

distribution on 1/@,; © 2, from (A.1) we have E(1/@ ,) = 2.0 = a/6, anda small value of 6 is associated with a larger variance. Accordingly, we set 6 = 2, or dg = 4 to obtain a large variance and a = 4, or a = 8 to obtain the desired mean. We assumed 012 ~ N(0, 1) to take a neutral stance on its sign and to set a fairly large variance. For the simulation, we set a burn-in of 10,000 and keep 40,000 generated values. A large number of iterations is chosen because of the complexity of the model.

descriptionView Paper arrow_downwardDownload

Analysis of Panel Data

by ali esem

descriptionView Paper arrow_downwardDownload

Economic Forces and the Stock Market Author(s): Nai- AM All use subject to JSTOR Terms and Conditions Economic Forces and the Stock Market

by Santiago iguaran

descriptionView Paper arrow_downwardDownload

Problems with Econometric Models: Heteroscedasticity, Autocorrelation & Multicollinearity

by Mwirigi Kiula

descriptionView Paper arrow_downwardDownload

An Analysis on the impact of Social Media on College Students Academic Performance

by Ferdinand 施南竹 Ilagan

College Students’ Academic Performance | ILAGAN gender, spent time browsing social media in a week. The model specification, definitions

and check using gretl, the following results were obtained: In the given OLS regressed model data, in order to determine if a variable is obtain the regressed model. By putting all the dependent and independent variable to Stata

In the VIF test, all the variables should be less than 10 for the model to experience the Inflation Factor (VIF) test. The data was computed using Stata, and is shown below: experiences mutlicollinearity, the Ordinary Least Squares Model (OLS) will still be linear and An analysis on the impact of Social Media on

model is shown below: y omitting the variables and regressing the model again. The results of the final regression Since irrelevant variable bias was violated, it is not always possible to drop th model. The R’ appeared to be having lesser goodness to fit as 20.68% are being explained in

The results from the VIF test demonstrated that there are no sign of dangerous

An analysis on the impact of Social Media on

Cameron & Trivedi's decomposition of IM-test An analysis on the impact of Social Media on College Students’ Academic Performance | ILAGAN

descriptionView Paper arrow_downwardDownload

E-Waste in Transition - From Pollution to Resource

by Florin Constantin MIHAI

E-waste management is a serious challenge across developed, transition, and developing countries because of the consumer society and the globalization process. E-waste is a fast-growing waste stream which needs more attention of... more

1. E-Waste Management in transition and developing countries

Figure 1. E-waste management interactions in a transitional stage

ment is a global environmental concern. Governments and local authorities across the globe face serious challenges in order to collect, treat, recycle and dispose this fast growing waste stream in a safety manner for the environment and human health. The global interconnection: between developed and developing countries, national and regional analyses are furthe revealed in the book.

The financial support for this research was provided by the Australia India Strategic Research Fund Round 6, Department of Innovation Industry, Science and Research, Australia. 2 Advanced Materials Technology Department, CSIR-Institute of Minerals and Materials Technology, Bhubaneswar, India

Figure 1. Sampling locations in Zhejiang province.

Figure 3. Comparison of PCDD/Fs- and PCBs-TEQ in human body fat from various countries [17-19, 23, 24, 28-30].

Figure 5. Comparison of PCBs concentrations (ng/g lipid) in children’s blood between polluted areas (LQ) and control areas (LY and TT) [39]. Different from the results of PCBs, PCDD/Fs levels in children’s blood of the heavily pollute area LQ (206 + 157 pg/g lipid) were shown to be higher than the moderately polluted area T (160 + 102 pg/g lipid), but lower than the control area LY (282 + 261 pg/g lipid). Integrate analysis of the data for the three areas was performed, and the average PCDD/Fs level | children’s blood was 208 + 172 pg/g lipid (54.4-784 pg/g lipid) (Figure 5). Our previous stuc also revealed that PCDD/Fs levels in food (primarily fish and egg) in LY were about 3- to. fold of those in LQ. In 1960s, PCB-Na was abundantly produced and extensively applied | control schistosomiasis in LY. But during the production of PCP-Na, a large amount of PCDI Fs was generated. PCDD/Fs are persistent and can bioaccumulate through food chai eventually enter human body [44]. These results in our studies indicate there still exists gre health risk to the environment and population due to historically widespread application | PCP-Na in LY.

elements. Materials such as gold and palladium can be mined more effectively from e-waste compared to mining from ore [28]. By contrast, e-waste contains PBDEs, which are flame retardants that are mixed into plastics and other components. Circuit boards found in most of the electronic devices may contain arsenic (As), cadmium (Cd), chromium (Cr), lead (Pb), mercury (Hg), and other toxic chemicals. Typical printed circuit boards treated with lead solder in electronic devices contain approximately 50 g of tin-lead solder per square meter of circuit board [7]. Obsolete refrigerators, freezers, and air conditioning units contain ozone depleting Chlorofluorocarbons (CFCs). The prominent materials such as barium, cadmium, copper, lead, zinc, and other rare earth metals are contained in end-of-life (EOL) cathode ray tubes (CRTs) in computer monitors, and televisions. For example, items such as leaded glass provide protection against X-rays produced in the picture projection process in CRTs [6]. The average lead in CTR monitors is 1.6-3.2 kg. Thus, the US and other developed countries in the EU and Japan have banned the disposal of cathode ray tubes in landfills because of their toxic charac- teristics. A critical challenge in designing and developing strategies to manage e-waste is the changing composition of the many constituents due the advancement of technology, particu- larly in the electronic components [24]. It is against this background that e-waste recycling and disposal methods ought to keep pace with the changing composition of e-waste. Several factors influence the composition of e-waste, including economic conditions, availability of a reuse market, and infrastructure of the recycling industry, waste segregation programs, and regulation enforcement. Figure 1 illustrates the distinctive materials ina WEEE. E-waste generated from the different diverse sources is normally collected as a whole unit or sub-unit of functional equipment. In many instances across the globe, whole units of e-waste

Figure 2. Flow of the take back system in Japan. Source; Adapted from [58].

Figure 1. International and national legal framework in terms of waste in Mexico.

Figure 2. International agreements which Mexico has signed in environmental terms. environment and where management of electronic waste is involved.

Figure 4. Environmental impacts and human exposure in life cycle of EEE.

Figure 5. Conceptual Life Cycle of Electrical and Electronic Equipment.

Figure 6. Actors involved in the value chain of electronic waste.

Figure 7. Stakeholders in the life cycle of e-waste.

Figure 8. Principles for the recovery of disused electronic equipment equipment market and sloughed them or incorporate new devices to prolong its life. One ¢ the problems associated with the recovery of functions is that the actors involved refurbishin EEE to incorporate them back into the market, as repairers or waste pickers, not considere parts of EEE as potentially hazardous substances since the usually they handled withot caution, with low technological support without concern that accumulate debris discar obsolete equipment or parts. In this sense it is important to consider the Swiss model fc electronic recycling in which they propose three principles, recovery function, material an energy. (Figure 8).

Figure 10. Estimed annual generation of electronic waste in Mexico. strategies for managing WEEE's and determination of impacts associated with it is important to consider a whole range of factors related to both the product and the consumer and post- consumer keeper’s management. In the case, electronics can be analyzed and evaluated environmentally in each stage of the life cycle. From a consumer perspective we might consider aspects such as the level of environmental awareness, consumption habits, socioeconomic status, disposal practices, etc. Also important to consider that a management model will be influenced by population issues such as: economic status, geographic location, cultural level, etc.

Figure 11. Flow for the recovery of electronic waste in Mexico. WEEE'’s. Informal recycling is an issue that must be addressed by the environmental implica- tions and associated health this common practice in our country.

Electronic waste, or e-waste, is an emerging problem with developed nations as with developing nations. In the absence of proper collection and disposal systems, aware- ness, and proper regulations, the problem is rather more acute in developing nations. These wastes are environmentally hazardous on one hand and valuable on the other. They contain substantial amount of metal value, including precious metals. Personal computers are the biggest contributors to e-waste, followed closely by televisions and mobile phones. The growth in their consumption pattern indicates a manifold increase in the volume of e-waste and calls for immediate attention to the manage- ment of e-waste in general and their recycling and reuse in particular.

Figure 3. Material composition of a typical TV.

Figure 4. Material composition of a typical mobile phone.

State-wise E-waste Generation in India (Tonnes/year) Figure 5. State-wise e-waste generation in India. The results of a field survey conducted in Chennai, a metropolitan city of India, to assess the average usage and life of the personal computers (PCs), televisions (TVs), and mobile phones demonstrated that the average household usage of the PC ranges from 0.39 to 1.70 depending on the income class [49]. Although the per-capita waste production in India is still relatively small, the total absolute volume of wastes generated is gigantic, and it continues to grow at an alarmingly fast rate. The growth rate of mobile phones (80%) is very high compared to that of PCs (20%) and TVs (18%). The public awareness on e-wastes and the willingness of the public to pay for e-waste management, as assessed during the study, based on an organized ques- tionnaire revealed that about 50% of the public are aware of environmental and health impacts of EOL electronic items. The willingness of the public to pay for e-waste management ranges from 3.57% to 5.92% of the product cost for PCs, 3.94% to 5.95% for TV and 3.4% to 5% for the mobile phones [50].

Figure 6. E-waste generation in India: Past and forecasts for the future. There are over 75 million mobile users and the number has increased to 200 million as of 2008 [57]. An estimated 30,000 computers become obsolete every year from the IT industry in Bengaluru alone [58]. India has about 15 million computers and the base is expected to grow to 75 million computers by 2010 since the life cycle of a PC has come down to 3-4 years from 7 to 8 years a few years back, and the segment is suffering from an extremely high obsolescence has jumped from 50,954 in 2002-2003 to 4,31,834 in 2005-2006 having registered an astonishing growth rate of 143% in 2005-2006 [48,52]. The overall PC sales in 2012-2013 considerably slowed down and the sales figure are well below the expectations. The overall sales figures touched 11.31 million in 2012-2013, registering a growth of 5% over the last fiscal. Desktop PCs continued to dominate the sales proceedings contributing around 60% of the sales although it is somewhat lesser than last year's contribution of 63%. Notebook sales posted a muted growth rate of 10% in 2012-2013 compared to the 22% rate in the previous year. Tablet PCs witnessed a massive growth rate of 424%. The sales for 2012-2013 stood at 1.9 million units as against 0.36 million units in 2011-2012 [53]. Sixty-five cities in India generate more than 60% of the total e-waste generated in India. Ten states generate 70% of the total e-waste in India [54]. Maharashtra ranks first followed by Tamil Nadu, Andhra Pradesh, Uttar Pradesh, West Bengal, Delhi, Karnataka, Gujarat, Madhya Pradesh, and Punjab in the list of e-waste gener- ating states in India (Figure 6). According to forecast, based on a logistic model and material flow analysis [55], the volume of obsolete PCs generated in developing regions will exceed that of developed regions by 2016-2018. By 2030, there would be two obsolete PCs in the developing world for every obsolete PC in the developed world. Similar forecasts have been arrived independently [56]. The advent of LCD, plasma, and larger screens has changed the way India views television and this has translated into phenomenal growth in sales, resulting in a considerable surge in rate of disposal as well.

Figure 7. Union Miniere Company's copper-smelting flowsheet for recycling of scrap IC board [14]. p: density (10° kg/m), o: electrical conductivity of material (10*/Om). The separation of metallic components through magnetic and eddy current separators are in vogue, wherein, ferrous components are separated, aided either by a permanent magnet or electromagnet, while metals such as aluminum and copper from non-metallic materials are separated in eddy current separator. Table 5 shows the materials that can be separated by eddy current separator. The main separation criteria is 0/9 [75]. On the basis of information provided by the Union Miniere Company [14], Figure 7 presents a copper-smelting flowsheet for recycling of scrap IC boards that is ideally carried out in a primary copper smelting plant, however, such facilities are not well-established in most parts of the world. Thus, removal of the non-recyclable materials (e.g., epoxy resin and fiber glass) from the IC board to enhance the value of recyclable material is preferable since post-separation provides higher metal concentration in lesser volume, thereafter the enriched metal content can then be sold and transported to an appropriate recycling facility for further processing [14]. transported to an appropriate recycling facility for further processing [14].

Figure 8. Huei-Chia-Dien Company's physical separation flowsheet for recycling of scrap IC boards [14]. Generally, this type of separation plant comprises of a series of physical treatment units devoted to processes such as crushing, grinding, screening, magnetic separation, air classifi- cation, eddy-current separation, electrical-conductivity separation, etc., wherein varied metal fragments of various size and content are obtained, depending on the separation technique and units deployed. The varied metal fragments, except iron, usually contain multiple types of metals, thus, identifying appropriate recycling markets for such mixed metal fragments is imperative [14]. There being no necessity of either water or chemical additive in the processing method, there is no wastewater-associated pollution issue, however, special attention should be provided with respect to dust and noise pollution. The low capital and operational cost in a physical separation plant for IC board recycling, being much less compared with a copper- smelting plant, is undoubtedly an added advantage of immense significance. On the basis of information provided by Huei-Chia-Dien Company, Taiwan [14], Figure 8 presents a physical separation flowsheet for the recycling of scrap IC boards.

Figure 9. Process flow chart for the technology developed for precious metals at CSIR-NML, Jamshedpur [77]. Processing technology has been successfully developed for the recycle and reuse of e-waste at Council of Scientific and Industrial Research—National Metallurgical Laboratory (CSIR-NML), Jamshedpur, India, in which metal bearing e-waste components were shredded and pulver- ized at the initial operation stage. Subsequently, the metals are separated from the plastics in the particulate mass, adopting a series of physical separation processes. The process does not require much specialized and sophisticated equipment for processing of waste PCBs, since the said equipment and machinery required are readily available, however, its efficiency, espe- cially with respect to commercial viability needs to be further worked upon [76]. The natural hydrophobicity of non-metallic constituents is effectively exploited by a flotation process and a continuous operation at plant level can reasonably be expected to minimize the loss of ultrafine metal values to a negligible level. The operation is simple and the overall

Pyrometallurgical processing techniques, including conflagrating, smelting in a plasma arc furnace, drossing, sintering, melting, and varied reactions in a gas phase at high temperatures for recovering non-ferrous metals, as well as precious metals from e-waste, happens to be the conventional method deployed in the past two decades, wherein, the crushed scraps are liquefied in a furnace or in a molten bath to remove plastics and in the process, the refractory oxides form a slag phase together with some metal oxides. In the precious metals refinery setup, gold, silver, palladium and platinum are recovered. The anode slime from the copper electrolysis process is subjected to pressure leaching, followed by drying of the leach residue and the same after addition of fluxes is smelted in a precious metals furnace, leading to the recovery of selenium. The remaining material, primarily silver, is cast into a silver anode, subsequently when subjected to a high-intensity electrolytic refining process, a high-purity silver cathode and anode gold slime are formed while leaching of anode gold slime leads to precipitation of high-purity gold, as well as palladium and platinum sludge. Figure 10 shows the precious metals recovery process. Recovery of precious metals from electronic scraps factually is the key to its commercial exploitation by the recycling industry, for profiteering, in the backdrop of the fact that e-scrap contains more than 40 times the concentration of gold content in gold ores found in the US [79], which is almost one-third the precious metal recovered in e-waste processing. The extraction of the precious metal is carried out by the well-established techniques that are discussed in detail in various articles [80-83]. Various methodologies such as pyrometallurgy, hydrometallurgy, and bi-hydrometallurgy technologies are analyzed for the recovery of gold and also the evaluation of recovery efficiency of gold from e-waste has been reviewed [84].

Figure 12. Schematic diagram for the Rénnskar Smelter [31]. Umicore published [30, 87] its precious metals refining process at Hoboken, Belgium, which is primarily focused on the recovery of precious metals from e-waste. Various industrial wastes and by-products from other non-ferrous industries (e.g., drosses, matters, speiss, anode slimes), sweeps of precious metals and bullions, spent industrial catalysts, as well as consumer recyclables such as car exhaust catalysts or PCBs are acceptable for the integrated metals smelter and refinery process. The plant treats around 2,50,000 tons of varied wastes per annual, Figure 11. Schematic diagram for the Noranda Smelting Processing [85].

Figure 13. Flowsheet for Umicore's integrated metals smelter and refinery [30]. out of which electronic waste presently comprises up to 10% of the feed [30]. It is the world's largest precious metals recycling facility with a capacity of over 50 tons of PGMs, over 100 tons of gold, and 2400 tons of silver [88]. The first step in the precious metals operations (PMO) is smelting by using an IsaSmelt furnace. Plastics or other organic substances that are contained in the feed partially substitute the coke as a reducing agent and energy source. The smelter separates precious metals in copper bullion from most other metals concentrated in a lead slag, which are further treated at the Base Metals Operations (BMO). The copper bullion is subse- quently treated by copper-leaching and electrowinning and precious metals refinery for copper and precious metals recovery. The Bace Metale Onerations process bv-nrodiicts from the PMQ). The main pnrocessine stens

The flowchart (Figure 14) shows two methods of recycling scrap mobile phones developed in Korea [38]. The first method (process I) involves shredding of waste PCBs and shipment to a copper smelter. The second method (process II) comprises of shredding, conflagration, melting or converting to copper alloy containing precious metals, and subsequent refining adopting the hydrometallurgical route. However, the systemic operation of recycling for e-waste processing operations in Korea does not in true sense function effectively since the majority of waste mobile phones collected are exported or conflagrated and landfilled, while only 2.5% of the waste mobile phones collected are actually processed for recycling. A pilot plant to recover cobalt from spent lithium-ion batteries of waste mobile phones is under operation, taking into account the high-valuation of cobalt. Figure 14. Flow sheet for the recycling of metal values from waste mobile phones in Korea [38].

Figure 1. Core strategies towards increasing the use of secondary raw materials, Source: Own illustration

Note: These data were determined by the authors.

Note: These data were determined by the authors. n, number of soil samples. Table 2. Concentrations of PCP/PCP-Na and DDT (ng/g dw) in soil of five regions.

Note: These data were determined by the authors. n, number of soil samples.

Table 4. PCBs concentrations (pg/g dw) and TEQ (pg/kg) in late rice, egg, crucian in different areas (n = 5).

Note: These data were determined by the authors. n, number of soil samples. Table 6. Concentrations of PCP/PCP-Na and DDT (ng/g dw) in crucian in several YH areas in HZ.

Many researchers have established that a clear definition of e-waste is needed due to rapid technological changes and enhancement, which are shortening the lifespan of the electronic products [8-10]. To date, the widely accepted definition in different e-waste studies is by the EU WEEE Directive, which defines e-waste as “Electrical or electronic equipment (EEE) which is Anelectrical and electronic product can be classified as a product that contains a printed circuit board (PCB) and uses electricity. Much has been written about the e-waste problem, yet the definition of the term "electronic waste" is quite complex to define. Referring to scholarly literature on the topic, there is, as yet, no standard definition, as every country has its own definition of e-waste. The questions that arise, therefore is: What is to be called e-waste? Any electronic or electrical appliances, which are obsolete in terms of functionality? Products that are operationally discarded? Or is it both? [10]. Table 1 gives a list of the different definitions of e-waste.

Disposal of e-waste is mainly through landfilling. Most often, the discarded electronic goods finally end-up in landfill sites along with other municipal waste or are openly burnt releasing toxic and carcinogenic substances into the atmosphere. In developing and transition countries the disposal of e-waste in the informal sector is very rudimentary so far as the safe techniques employed and practices are concerned, resulting in low recovery of materials [38]. Table 3 presents a comparison of typical disposal systems in developed and developing countries. presents a comparison of typical disposal systems in developed and developing countries. E-waste management is different between developed countries and developing and transition countries. Developing and transition countries do not have guidelines and information campaigns on the fate of e-waste. Especially, less sophisticated disposal systems are used, from open burning and dumping to uncontrolled landfill sites, which pose significant environmental pollution and occupational exposure to e-waste-derived chemi- cals [31]. Serious challenges in the disposal of e-waste were analyzed across developing countries such as Brazil [19], China [42], and India [43], outlining the difficulty to implement/ enforce existing regulations and clean technologies backed by lack of capacity building and awareness. In contrast, developed countries have devised sophisticated disposal schemes and high-cost systems, which are less hazardous to handle waste. However, a comprehen- sive overview of the situation is constrained by the availability of data. This means that the differences in the socio-economic and legal contexts between typical developing and developed countries’ scenarios limit e-waste management in developing and transition countries. The regulations that guide the disposition of e-waste in developing countries is mostly fragmented and lack monitoring, while in developed countries the regulations are stringent and there is effective monitoring [36].

Table 1. E-waste sources and their health effects. 4.9. Health and environmental impact of e-waste EOL of electrical and electronic equipments comprise numerous components, many of whicl are inherently hazardous and highly toxic in nature, which if not arrested through scientifical]; sustainable recycling and disposal, can lead to a disastrous impact on life, environment, anc climate as well. Certain examples of sources of e-waste and their related adverse health impact are listed in Table 1 [23]. However, if handled in a controlled environment and disposed-of adopting safe and sustainable methodology, these e-wastes provide immense value additio1 and new product cycle, driving great economic prospect, without posing risks to life, envi ronment, and climate. However, haphazard recycling and disposal of e-waste by the unor ganized sector without access to adequate technology and resources, guided by profit-onl} motive can have damaging consequences to inhabitants and the environment, including bu not limited to the workforce engaged in this trade, groundwater pollution, etc., especially or account of highly toxic release into the soil, air, and ground water [23].

Given the diverse range of materials found in WEEE, it is difficult to give a generalized material composition for the entire waste stream. However, most studies examine five categories of materials: ferrous metals, non-ferrous metals, glass, plastics, and others. Figure 1 shows the material fractions in e-waste [2]. Metals are the major common materials found in e-waste representing about 60%. Plastics are the second largest component by weight representing about 15%. Figures 2—4 shows the material composition of a personal computer [25, 26], followed by television sets [27] and mobile phones [28].

The recycling/recovery of valuable substances by industries in the organized sector with access to requisite technology and manpower is carried out in protected environment, adopting adequate preventive methodology to minimize damage to life and environment. The merit of a focused approach by the stakeholders factually complements the efficacious recovery of metals, including rare and precious metals present in traces, aided by advanced process technology, wherein the processing capacity or volume plays a pivotal role in contributing to the viability aspect, keeping in mind the high cost of capital investments for infrastructure

The content or substances in cellular phone are variable to some extent, based on the model and its manufacturer, with no fixed formula or list of contents applicable as such, thus, the list of substances in an average mobile phone may also be misleading since varied substances might be used as additives in very minimal quantities or traces by different manufacturers in the production of microelectronic components. However, the general composition of cellular phones and other small electronic goods as well, is identical in nature. Table 6 presents the fractional composition of a modern cell phone [89]. Recovering metals of higher percentage concentration like copper and metals of precious value or worth like gold, palladium and silver is factually the underlying objective for metal recovery from EOL or obsolete cellular phones and aluminum or magnesium cases of cellular phones wherever applicable, contribute further to value addition or generation through its recycling.

4.2. Bundling the instruments into core strategies Looking at the different instruments it becomes clear that there is no lack of ideas, the key challenge is obviously the implementation phase. The following table provides a general overview over the investigated instruments and the evaluations of quantity effects (+++ high quantity effects) and their feasibility (+++ generally high feasibility) carried out in the process. There is a clear trade-off between these two analytical dimensions: Instruments with poten- tially high quantitative effects often seem rather unrealistic to implement. Feasible instruments on the other hand let expect so low effects that the transaction costs of policy developments maybe equal or higher, preventing e.g. the change of regulatory frameworks. naybe equal or higher, preventing e.g. the change of regulatory frameworks.

descriptionView Paper arrow_downwardDownload

Megaproject Planning and Management: Essential Readings, vols. 1-2

by Bent Flyvbjerg

2014

Megaproject Planning and Management: Essential Readings contains the seminal articles from the growing body of research on megaproject planning and management along with an original introduction by the editor, Bent Flyvbjerg. The leading... more

descriptionView Paper arrow_downwardDownload

Titman wessel - determinants of cs 1988

by Geteye Addis

descriptionView Paper arrow_downwardDownload

Investor protections and economic growth

by Jamal Ibrahim Haidar

Using objective measures of investor protections in 170 countries, I establish that the level of investor protection matters for cross-country differences in GDP growth: countries with stronger protections tend to grow faster than those... more

descriptionView Paper arrow_downwardDownload

Neoliberalism and the geographies of marketization: The entangling of state and markets

by Kean Birch

Increasingly, governments are experimenting with ways to provide public goods by involving the private sector in the planning, financing, building and operating of a range of services, facilities, infrastructure, etc. In the geographical... more

descriptionView Paper arrow_downwardDownload

Analisis Regresi dan Uji Asumsi Regresi Linier

by Rezzy Eko Caraka

Analisis regresi linier sederhana dapat digunakan untuk menentukan persamaan regresi yang menunjukkan hubungan antara variabel dependen yang ditentukan dengan satu variabel independen. Persamaan variabel yang diperoleh dari proses... more

descriptionView Paper arrow_downwardDownload

Asset Pricing: (Revised

by Wayne Sun

Hansen—Jagannathan bound declines. At the mean return correspondin:

Figure 5.5. Decomposition of any discount factor m = x* + we + 1

The S matrix is the long-run covariance matrix of the moments,

Figure 15.1. Average excess returns vs. betas on CRSP size portfolios, 1926-1998. The lin
gives the predicted average return from the time-series regression, ECR’) = BE(R®").

Figure 17.1. Payoff diagrams for simple option strategies.

Figure 17.4. Arbitrage bounds for a call option.

-lugging in the mean from (19.28) and the variance from (19.29)

Table 15.1. Parameter estimates and standard errors

Table 15.2. x°* tests that all pricing errors are jointly equal to zero

Table 15.4. Power. Probability of rejection for x* statistics under the null that the CAPM is fals
and the true means of the decile portfolio returns are equal to their sample means

ble 15.3. Size. Probability of rejection for x statistics under the null that all pricing errors are zer
Table 15.4. Power. Probability of rejection for x* statistics under the null that the CAPM is fals
and the true means of the decile portfolio returns are equal to their sample means

Table 15.5. Monte Carlo and block-bootstrap evaluation of the sampling variability of parameter
estimates b and xr

Ri-+1+k indicates the k-year return. Standard errors in parentheses use GMM to correc
for heteroskedasticity and serial correlation. Sample 1947-1996.

Table 20.2. Long-horizon return forecasts

le 20.3. Variance decomposition of value-weighted
NYSE price/dividend ratio

Table 20.4. Estimates of log excess return and log dividend/price ratio regressions,
using annual CRSP data
is the difference between the log value-weighted return and the log treasury bill rat
‘he estimates are of the system

Table 20.5. Mean-reversion using logs, 1926-1996

Table 20.6. Mean-reversion using gross returns, 1926-1996
r denotes the difference between the gross (not log) long-horizon value-weighted NYSE
return and the gross treasury bill return.

Table 20.7. Mean-reversion in postwar data

Table 20.9. Forecasts based on forward-spot spread

2 20.10. Numerical example of the effect of measuremen
error in yields on yield regressions

tble 20.12. Average monthly returns from reversal and
momentum. strateoies

Table 21.1. Means and standard deviations of simulated and historical
data
Table 21.1. Means and standard deviations of simulated and historical

Table 21.2. Long-horizon return regressions

descriptionView Paper arrow_downwardDownload

The Determinants of Capital Structure: Evidence from China

by Samuel Gui

2002, SSRN Electronic Journal

This paper employs a new database, which contains the market and accounting data from more than 1000 Chinese listed companies up to the year 2000, to document the characteristics of these firms in terms of capital structure. As in other... more

wow

Note: "+ " means that leverage increases with the factor. "-" means that leverage decreases with the factor.

Notel: Book long-term debt ratio, LD, is defined as long-term debt divided by long-term debt plus book

Table 3 The Correlation Matrix of Leverage and Independent Variables for Chinese Listed Firms

Table 5 OLS Analysis Results on Total Liabilities Ratios for Chinese Listed Companies

descriptionView Paper arrow_downwardDownload

The Validity of Capital Asset Pricing Model (CAPM) and Arbitrage Pricing Theory (APT) in Predicting the Return of Stocks in Indonesia Stock Exchange

by ZAINUL KISMAN

Financial experts have developed two approaches to measure the required return of stock, those are the Capital Asset Pricing Model (CAPM) and Arbitrage Pricing. Theory (APT). CAPM explains that stock return is the sum of the risk free... more

Figure 1. Flow of this Research analytical and framework of study was made to show the flow this research (see Figure-1). In the first, we take LQ45 firms in IDX.

Table 3.1. The results of calculation of regression coefficients, T-test and Significance-value Source : Data Processed with SPSS 17.

Table 3.2. The results of calculation of coefficient of determination Source : Data Processed with SPSS 17.

Source : Data Processed with SPSS 17. Table 3.3. The results of calculation of regression coefficients, T-test and Significance-value

Source : Data Processed with SPSS 17. Table 3.4. Result of the F-test Value 1) Gross Domestic Product (GDP). Table 3.5. The Result of Calculation of Coefficient of Determination To determine whether there is simultaneously influence of independent variables (GDP and Interest rate) on the dependent variable (stock return), this study used F-test (Gujarati, 2008; Hill et al, 2001).

Source : Data Processed with SPSS 17. market the higher the stock return and the market more confidence on the prospects of the company.It will be an attractive for investors to buy it. And demand for these shares will go up, then concluded that the push the stock price rises. It can be higher the excess return will contribute to the higher return of the stock, or vice versa. These findings support the results and Muhammad M the excess return stock return. Based of studies of Candra Gancar Premananto adyan doing research in 2004 showed that is positive and significant effect on the on the results of this study excess market returns can be used as predictors in predicting stock returns. Of course besides that, the investors in predicting return also pay attention to t he performance of the company (issuer, including its risk) and macro-economic condition.

descriptionView Paper arrow_downwardDownload

Soft Spying: Leveraging Globalization as Proxy Military Rivalry, International Journal of Intelligence and Counterintelligence, Vol. 28, No. 1, Spring 2015

by Dr. Matthew Crosston

descriptionView Paper arrow_downwardDownload

Stasioneritas, Akar Unit, & Kointegrasi (Pengantar Ekonometrika Time Series)

by Arif Rahman Hakim

Time series merupakan salah satu jenis data yang sering kita jumpai. Karakteristik data time series melibatkan lebih dari satu titik waktu dengan unit analisis dapat berupa kota, kabupaten, perusahaan, negara dan sebagainya. Sedangkan... more

descriptionView Paper arrow_downwardDownload

Introduction to Econometrics Third Edition Contents

by Lubna Elshikhi

descriptionView Paper arrow_downwardDownload

Determinants of Capital Structure and Adjustment to Long Run Target: Evidence from UK Company Panel Data

by Franciele Daciê

The major objective of this paper is to provide more insight into the empirical determinants of target capital structure of firms and the adjustment process toward this target. This study extends the empirical research on the topic of... more

descriptionView Paper arrow_downwardDownload

BOOK BIOESTADISTICA PARA NO ESTADISTICOS

by jota te

Estadistica para profesionales de las ciencias sociales, sin bases mayores en matemáticas

descriptionView Paper arrow_downwardDownload

Testing the null hypothesis of stationarity against the alternative of a unit root

by Yongcheol Shin

2001

We propose a test of the null hypothesis that an observable series is stationary around a deterministic trend. The series is expressed as the sum of deterministic trend, random walk, and stationary error, and the test is the LM test of... more

Size of 7, and 4,, 5% level, iid errors. integer[12(7'/100)'/*]. These choices of / follow Schwert (1989) and other recent simulations.

Size of 4, and 4,, 5% level, AR(1) errors.

Power of 4, and 4,, 5% level, iid errors.

7H, and 7, tests for trend stationarity applied to Nelson—Plosser data. Table 5

descriptionView Paper arrow_downwardDownload

Complex Effect of Working Capital

by bagus anu

The main aim of this article is to examine the effect of working capital on profitability of Indian firms. We collected data about a sample of 263 non-financial BSE 500 firms listed at the Bombay Stock (BSE) from 2000 to 2008 and... more

descriptionView Paper arrow_downwardDownload

What Do We Know About the Profitability of Technical Analysis?

by James V Carter

2007, Journal of Economic Surveys

The purpose of this paper is to review the evidence on the profitability of technical analysis. The empirical literature is categorized into two groups, 'early' and 'modern' studies, according to the characteristics of testing procedures.... more

*Studies on equity (index) futures and options and foreign exchange futures are categorized into ‘stock markets’ and ‘foreign exchange markets’ studies, respectively. ‘Futures markets’ studies include studies on other individual futures markets or various groups of futures markets. >Through August 2004. Table 1. Number of Technical Trading Studies, 1960—2004.?

Table 2. Categories for Modern Technical Analysis Studies, 1988-2004. “The total number of modern studies is 95.

descriptionView Paper arrow_downwardDownload

Pengenalan Eviews dan Statistik Deskriptif

by Rezzy Eko Caraka

Eviews adalah suatu software yang berfungsi untuk menganalisis data, melakukan analisis regresi, dan melakukan peramalan dengan basis Windows. Dengan fasilitas-fasilitas yang tersedia di software ini, pengguna dapat dengan mudah membangun... more

descriptionView Paper arrow_downwardDownload

DATA PANEL : TEORI DASAR DAN APLIKASI DI STATA Oleh : Akbar Suwardi

by Akbar Suwardi

descriptionView Paper arrow_downwardDownload

Contoh Soal & Jawaban SPSS

by Rizky Dwi Yulianto

Contoh Soal & Jawaban SPSS. soal soal ekonometrika saat semester 5 kuis ekonemtrika

descriptionView Paper arrow_downwardDownload

Beyond sacrificial harm: a two-dimensional model of utilitarian psychology

by Guy Kahane and

2018, Psychological Review

Recent research has relied on trolley-type sacrificial moral dilemmas to study utilitarian versus nonutili- tarian modes of moral decision-making. This research has generated important insights into people’s attitudes toward instrumental... more

descriptionView Paper arrow_downwardDownload

The stocks at stake: Return and risk in socially responsible investment

by Robin Dorre

We relate US portfolio returns, book-to-market values and excess stock returns to different dimensions of socially responsible performance. We find that socially responsible investing (SRI) impacts on stock returns by lowering the... more

Cross-section regressions of excess stock returns, July 1992-June 2006 This table reports the results on the Fama and MacBeth (1973) regressions. Variables are described in Table 1. Dummy Industry Y (N) indicates whether 39 industry dummi have (have not) been included in the specification. T-statistics are reported below the coefficients. " 10% significance “ 5% significance “” 1% significance. Table 3

Table 1 ~ This table reports summary statistics on the variables used in the cross-sectional regressions. The summary statistics of the SRI variables are reported in Panel C. Excess return is a stock’s monthly return net of the risk free rate. Beta is a stocks post-ranking Beta calculated following Black (1972). Log(Size) is the logarithm of company market capitalization at the end of month t — 1. Log(B/M) is the logarithm of the book-to-market ratio at the end of month t — 1. Return is the simple average of returns during the

In Panel A we estimated for all portfolios in a system GMM framework the regression: R(t) — RF(t) = « + B[RM(t) — RF(t)] + s SMB(t) + h HML(t) + m MOM(t) + e(t). T-statistic: are adjusted following Ferson and Foerster, 1994 by a factor T/(T-P), where T is the number of time periods and P the number of parameters. We also test the over identifyins restriction generated by the remaining orthogonality conditions when we set one of the parameters to zero for all of the portfolios. For panel A Hansen (1982) shows this tes is distributed as y?(6) under the null since in the restricted case the number of instruments is 36 and the number of parameters is 30 so that the number of over identifyin; restrictions is 6. In panel B this test is distributed as y7(12) under the null since in the restricted case the number of instruments is 72 and the number of parameters is 60 s« that the number of over identifying restrictions is 12. J-stat * and ® denote J-stats from 2-Stage and Iterated GMM, respectively. P-values are reported between brackets. Th: covariance matrix is computed following Newey and West (1994). The Bartlett spectral density kernel estimator is used to make sure that the covariance matrix is positive semi-definite. For a description of the variables we refer to Table 1. "10% significance ” 5% significance ” 1% significance. + Not available due to the fact that the iterated estimation procedure did not converge. Carhart regressions, equally weighted KLD portfolios, July 1992-June 2006 Table 2

Cross-section regressions of excess returns on employee relations screens This table reports the results on the Fama—MacBeth regressions. Variables are described in Table 1. Dummy Industry Y (N) indicates whether 39 industry dummies have been (have not been) included in the specification. Empstr. A: strong union relations. Empstr. C: Cash Profit Sharing. Empstr. D: Employee Involvement. Empstr. X: Other Strenghts. Empcon A: Poor Union Relations. Empcon. B: Safety Controversies. Empcon C: Workforce Reductions. Empcon D: Pension/Benefits Concern. Empcon. X: Other Concern. T-statistics are reported below the coefficients. ” 10% significance 5% significance “” 1% significance. Table 4

This table reports the results of the pooled book-to-market regressions. The vari- ables are described in Table 1. In column (1) we report results for the specification including time dummies, clustering standard errors by firm. Dummy Time Y (N) indicates whether 14 year dummies have (have not) been included in the specifi- cation. In column (2) we report results for the specification where we cluster by both year and firm. ™ 5% significance. 1Y cionificance Pooled book-to-market regressions 4.4. Robustness Table 5

descriptionView Paper arrow_downwardDownload

The Great Recession and (2007- 2008 Economic Crisis) - Definition, Start, Causes, Effects and Responses

by Ibrahim Rihan, MBA | PDDM

The Great Recession was a period of general economic decline observed in world markets during the late 2000s and early 2010s. The scale and timing of the recession varied from country to country.

descriptionView Paper arrow_downwardDownload

Principles of Econometrics, 4th Edition - Instructor's Manual

by Büşra Tüfek

Instructor's Manual
William E. Griffiths, R.Carter Hill, Guay C. Lim, Simon yunho Cho, Simone Si-Yin Wong

descriptionView Paper arrow_downwardDownload

Another look at measures of forecast accuracy

by Anne Koehler

2006, International Journal of Forecasting

We discuss and compare measures of accuracy of univariate time series forecasts. The methods used in the M-competition and the M3-competition, and many of the measures recommended by previous authors on this topic, are found to be... more

descriptionView Paper arrow_downwardDownload

A History of Market Performance from Ancient Babylonia to the Modern World

by R.J. (Bert) van der Spek and

2014, R.J. van der Spek, Bas van Leeuwen, Jan Luiten van Zanden (eds.), A History of Market Performance from Ancient Babylonia to the Modern World. London, New York: Routledge 2015 [2014]

This exciting new volume examines the development of market performance from Antiquity until the dawn of the Industrial Revolution. Efficient market structures are agreed by most economists to serve as evidence of economic prosperity,... more

descriptionView Paper arrow_downwardDownload

Independent Dummy Variable

by Rezzy Eko Caraka

Variabel Dummy adalah variabel yang merepresentasikan kuantifikasi dari variabel kualitatif. Misal: jenis kelamin, pendidikan, lokasi, situasi, musim, & kualitas. Jika data kualitatif tersebut memiliki m kategori, maka jumlah variabel... more

descriptionView Paper arrow_downwardDownload

The impact of government expenditure on economic growth: How sensitive to the level of development?

by Eric Lin

2010, Journal of Policy Modeling

Previous studies generally find mixed empirical evidence on the relationship between government spending and economic growth. In this paper, we re-examine the causal relationship between government expenditure and economic growth by... more

Notes: (1) Y, sg, y, g, and G denote real GDP, government share, real GDP per capita, real government expenditure per capita, and real government expenditure, respectively. (2) All variables are logged before testing. (3) A denotes the first-difference operator. (4) p-values are given in parentheses. Panel unit-root tests — level and first difference. Table 1

Notes: (1) Y, sg, y, g, and G denote real GDP, government share, real GDP per capita, real government expenditure pet capita, and real government expenditure, respectively. (2) The formulae for Z and Z are displayed in Eqs. (6) and (8). (3) The symbol K denotes the autoregressive lags. (4) All variables are logged and first-differenced before testing. “ Significance at 5% level. Panel Granger causality tests (Part I).

Panel Granger causality tests (Part II). Table 3 of government corruption. As noted above, corruption not only affects government efficiency and spending composition, but also hinders economic development. Consequently, the effect of government spending on economic growth would be lessened in a more corrupt environment. Moreover, corruption may induce more government spending on military spending (Mauro, 1998) and hence the effect of economic growth on government spending is likely to be larger in a more corrupt country.

Notes: (1) Y, sg, y, g, and G denote real GDP, government share, real GDP per capita, real government expenditure pe capita, and real government expenditure, respectively. (2) The formulae for Z and Z are displayed in Eqs. (6) and (8). (2 The symbol K denotes the autoregressive lags. (4) All variables are logged and first-differenced before testing. * Significance at 5% level. Panel Granger causality tests (Part III).

descriptionView Paper arrow_downwardDownload

Advanced International Trade: Theory and Evidence

by NGUYEN THI HONG NHUNG k48

descriptionView Paper arrow_downwardDownload

Islamic Banking and Economic Growth: Empirical Evidence from Malaysia

by haider ali

This paper examines the dynamic interactions between Islamic banking and economic growth of Malaysia by employing the Cointegration test and Vector Error Model (VECM) to see whether the financial system influences growth and growth... more

In this paper we use quarterly time series data (1997:1 — 2005:4) for the
variables Islamic Bank total financing (IBFinancing) as a represent of
‘financial sector’ and three variables representing ‘real economic sector’
namely real GDP (GDP), real gross fixed capital formation (GFCF),°
and trade activities that involve export plus import (TRADE). The data
are gathered from International Financial Statistics, IMF and Monthly
Statistical Bulletin of Bank Negara Malaysia.
“We use GFCF as a representation of investment as it is economic indicators of the
level of business activity that measure net new investment by enterprises in the
domestic economy in fixed capital assets during an accounting period.

Note: * significant at 1 % level, figures in parentheses are the standard errors.
Table 2: Johansen's test for the number of cointegrating vectors (VAR, 4lags)

Notes: *, ** significant at 1% and 5% level, figures in parentheses are the P-value
6. Conclusion
The results generally show that in the long-run, Islamic bank financing
is positively and significantly correlated with economic growth and
capital accumulation of Malaysia. In this regard, Islamic banking has
effectively played its main role as financial intermediaries that facilitate
the transmission of savings from surplus households to deficit
households. Thus, we can say that the current policies by Bank Negara
Malaysia to develop a comprehensive Islamic financial system in
Malaysia is considered as effective since financial development and

descriptionView Paper arrow_downwardDownload

Megaprojects and Risk: An Anatomy of Ambition

by Bent Flyvbjerg

2003

Back cover text: Megaprojects and Risk provides the first detailed examination of the phenomenon of megaprojects. It is a fascinating account of how the promoters of multi-billion dollar megaprojects systematically and self-servingly... more

descriptionView Paper arrow_downwardDownload

Government Spending and Economic Growth: Econometric Evidence from the South Eastern Europe (SEE

by Mega Mustika Sari

This paper provides further evidence on the relationship between economic growth and government spending. For the first time two different panel data methodologies have been applied to seven transition economies in the South Eastern... more

descriptionView Paper arrow_downwardDownload

Guide d'utilisation de STATA

by Dalila Chenaf- Nicet

Chapitre 3 : l'économétrie de base sous stata

descriptionView Paper arrow_downwardDownload

The Impact of Monetary Policy on Economic Growth in Iran

by Kuyooro Raphael

Evolution of money supply and gross domestic product are in a close relationship. This study examine the relationship between money supply and economic growth in Iran adopting ordinary least squares (OLS) technique and also uses data... more